find below few comments on the presentation of the Z/W cross section result in the CONF note (mostly style and comparison witih theory).

Abstract: L8: integrated --> inclusive or total (as opposed to differential, integrated gives the idea we are summing up differential which is not what we have done so far)

we believe that integrated is appropriate here, inclusive could still be differential as it means e.g. Z+X, total we have in the title and we distinguish it from fiducial, so we can either drop integrated or keep it, as we prefer.

L11: in the year 2010 -> in 2010

done

Introduction:

L19: it's quite bizzarre to qualify as "modern" NLO QCD Drell-Yan calculations that are available since more than 30 years ! Also, Bjorken x should be defined. Suggest:

Next-to-leading order (NLO) [2–4] corrections to the Drall-Yan process have been known since a long time, while in the past decade inclusive as well fully differential next-to-next-to leading order (NNLO) [5–9] calculations have become available and have been tested extensively at the Tevatron [reference to add]. A crucial ingredient of the theoretical calculations are parameterisations of the momentum distribution functions of partons in the proton (PDF). These have been determined in a variety of phenomenological analyses to NLO by the CTEQ [10,11] and NNPDF [12] groups and more recently also to NNLO by the MSTW [13], the ABKM [14, 15], HERA [16, 17] and the JR [18] groups. High precision measurement of W and Z cross section at LHC has the potential of improving the knowledge on PDF for low x, the proton momentum fraction carried by partons, in the following referred to as Bjorken x.

Agree that modern is debatable and take it off. We thought for the conference not to go into Tevatron history. The last sentence is somehow the outlook sentence of the paper summary. We prefer to have that discussion in the paper when we have differential distributions. We define now x Bj when its mentioned first.

I also suggest to stress in the introduction that this measurement is the first from Atlas enjoying a much reduced luminosity uncertainty thus enabling significantly more stringent tests than it was possible before, even w.r.t to Tevatron!

We comment on the lumi in l27. We have received comments in the other direction, i.e. asking us not to stress this too much.

Sec. 2.4 Analysis Procedure:

L113: An additional acceptance factor AW /Z , derived from simulation, --> derived from Monte Carlo calculations. (Detector simulation has no impact here, as formula 6 says explicitely)

ok

L118-122: The discussion of charge misid in the context of W+/- cross section determination should be complemented with some statement on the fact that this charge misidentification from MC is corrected from measurement from Data. At least this is what a reader would understand from reading the subsequent electron section. Also no mention at all of charge misidentification for muons is given. This may well be irrelevant but should be stated somewhere.

We say on l110, that C factors are corrected for data/MC differences. Added a sentence, that this is only relevant for the electrons, but not the muons.

Sec. 3.1 Electrons

W tag&probe method seems really jargon. A casual reader may not understand how a tag & probe may be implemented with only one lepton in the event. Please add some detail.

Ok, but we try to be brief.

The Etmiss fits are not so nice-looking (as usual, nothing particularly unexpceted). The paper says only that templates are varied and a systematics assessed this way. I would also add somehow that these variations are big enough to cover discrepancies seen at low Et miss in the QCD fits.

Ok

Sec 4.1 & 4.2

Discrepancies in Et(e), pt(mu) and mT distribution for W+/- are attributed repeatedly to missing NLO corrections (even though LO* pdf are used). The agreement is termed "reasonable" and the use of MC@NLO is mentioned as a way of evaluating a systematic uncertainty. So the reader is lead to think that MC@NLO provides a better description of ATLAS data, and then it will remain unclear why, for both plots and central values, Pythia had been used instead. On the other hand we now that MC@NLO is not definitely better than Pythia overall, and we convinced ourself that the above mentioned discrepancy have very little impact on fiducial (and total) cross section measurement. Suggest then to add somewhere a statement like: "The plots do not look extremely nice, but the assigned systematics cover the observed discrepancies in data vs MC distributions."

In our view this is exactly the statement we have there. We say that MC@NLO and ofther PDFs are "different", not necessarily better and that this is used for theoretical systematics. For the CONF note we are under very high time pressure and could not study possible improvements here.

Sec. 5.1 L344 : "forward Z is not included yet in this combination" (Guess we want this for the paper, to gain from the anticorrelation in acceptance uncertainty in the two measurement ?).

Yes, if it makes sense from the uncertainty point of view.

L 352: specify which are from FEWZ and which from ZWPROD in text and table caption.

We believe that is too much detail at this stage but will (have to) do it in the paper.

L 359: "deep inelastic scattering, mainly from Hera," -> "deep inelastic scattering, mainly from Hera, and collider data" (Not a real expert but I do believe that MSTW,ABKM at least, do include Tevatron vector bosons and jets)

The precision data in this x range is from DIS at HERA.

L360-361: The statement look a little bit ingeneous. I believe that remarkable agreement of measurements with DGLAP evolution functions has been reached elsewhere and before us. If the present test of DGLAP is more remarkable than past tests one should quantify and compare.

This measurement is certainly not the most stringent test of DGLAP to date. But it is a measurement in a slightly different kinematic domain than other measurements have done before. Therefore it is non trivial that this agreement is seen and it is due to pQCD and the pdf determinations. We agree the statement is general, and we know there have been impressive tests before..

Footnote 2: not sure to grasp the sense of this. First it looks new to me that NNLO prediction is higher than NLO. Read many times in the past that NNLO correction is reducing NLO cross section. See for example hep-ph/0312266 pag. 28 (ref. [9] in the conf note). Secondly it's obvious that, once higher precision calculation are available, and also PDF are calculated to NNLO , one should use NNLO predicitons to further constrain those. On the other hand, despite the great reduction in uncertainties from this measurement, we are still far from being able to tell NLO from NNLO. That would require a precision significantly smaller than the 3% NLO-NNLO difference quoted here, and we are still far from this. Could we simply skip this unnecessary footnote?

This was to justify our choice of NNLO over NLO, as e.g. the CTEQ set is not available in NNLO. The differences NLO-NNLO are small as stated (3%, which is documented in ATL-COM-PHYS-2010-695) and our measurement precision is on the same level already (~4%). If you look at the recent talk of Graeme Watt at pdf4lhc (Monday 7.3.) you see that the trend is that indeed NLO is lower than NNLO.

http://indico.cern.ch/conferenceDisplay.py?confId=127425. (confirmed for MSTW08, also seen at ABKM09). We could skip the footnote as indeed it is a comment only to the main text, we yet felt it brings interesting information, also on the scale dependence. In order not to go into too far detail, following your impression, we remove(d) footnote 3 which is on NNLO-NLO for the ratios.

Sec. 5.2 We do all sort of ratios but the W+/W- . It seems odd since we do have separate W+ and W- cross section right here in this document. Also about the comment on Line 385-386: is it indeed a comment on MSTW not being able to reproduce W+/W- ratio at the level of accuracy of the current ATLAS measurement? On the separate W+/Z and W-/Z we are talking about a 1 sigma (!!) effect that do not indeed warrant the above comment, in my opinion.

The W+/W- ratio needs some extra work to control cancelations in the systematics properly, but e.g. we have the W asymmetry coming out at the same time.

Fig. 8:

Substitude the line for a dot with error bars for "uncorrelated exp. + stat. uncertainty" item

done

The green ellipse: check numbers for W+ , it seems narrower than it should based on numbers in Tab 8, and for W- as it seems larger than it should.

In the caption there is a confuse statement about missing energy systematic (isn't it enough just saying missing energy systematics are treated as correlated?).

done

I would also list which systematic are instead taken as correlated and with which degree (lumi 100%, acceptance 100% (?), lepton id(100%)).

In the present version, we treated as 100% correlated only luminosity and acceptance (plus MET for W+ vs. W- plot). Clearly also lepton id. has some level of correlation. Now we preferred to take a conservative path and treat this as fully uncorrelated. Work is ongoing to understand better correlated effects on both muon and electron channels, that are now accounted for in the measurement systematics.

From the distance between the ellipse and JR point for W- vs W+ we should be able to claim that this PDF set id disfavored at some C.L., can't we? Especially keeping in mind that even theory is indeed a rather narrow ellipse.

As said before, we would at this time not make too strong statements about W+/W- before completing more work on their correlation. In fact, we found a rotation error of the ellipse as well, also triggered by your comment. Now the situation is less obvious.

Table 9: specify at which CL the PDF uncertainty is given (68% I guess), in literature it is often quoted a 90% CL range.

ok

Fig. 9 10: please make a substantial zoom on the x axis. Most of the space of the plot is currently occupied by white pixels.

DONE

regards,

Marco Rescigno

thank you!

-- MassimilianoBellomo - 07-Mar-2011